## Loading required package: dispRity
##        --- dispRity package ---
## This is the CRAN release version (1.6.0) of the package.
## For news, vignettes and future releases,
## visit https://github.com/TGuillerme/dispRity
## 
## Attaching package: 'beer'
## The following object is masked from 'package:datasets':
## 
##     trees
## Loading required package: Matrix
## Loading required package: coda

Number of dimensions required to capture at least X% of variance per group

This is the number of dimensions required per PCA contain X% of variance per group. The groups being either the full PCA (all - the classic way to measure the variance threshold) or each order (order) or family (family). The number of dimensions per group is measured either for the PCA with the size component removed (sizefree) and for the PCA with the size component (sizeform).

Variance per specific axis for the 37 first dimensions

Here I will display the scaled variance per axis for each group (as mentioned above). For example, for the full PCA, the results are pretty common and easy to grasp:

When considering all the data, the first three axis contain 95% of the variance (black vertical line) and PC1 to 6 for 99% of the variance (grey vertical line). For the PCA with size, the first axes contains 95% of the variance and the PC1 to 3 contains 99%. The rest of the variance is distributed pretty log-normally as expected.

We can then measure the same be measuring things per family or per order (i.e. within the PCA, how many axis are required to represent 95% of the variance of order/family X). The results vary much more depending on the order/family! For example, in Strigiformes, the first PC does not contain much variance at all compared to the other axes! Note that the order with just one bar at 100% just have one species so they don't count.